New Foundations for Physical Geometry Original URL: http://www.unil.ch/webdav/site/philo/shared/summer_school_2013/NYU.ppt Tim Maudlin NYU Physics & Philosophy of Time July 25, 2013

1. New Foundations for
Physical Geometry
Tim Maudlin
NYU
Physics & Philosophy of Time
July 25, 2013

2. A Puzzle and an Adage
Eugene Wigner famously commented on
“the unreasonable effectiveness of
mathematics in the natural sciences”.
In one form, this is a puzzle about why
mathematical constructions are so apt as a
medium of representation of the physical
world.

3. Numbers
The puzzle seems particularly acute with respect to
certain numerical representations, and especially
those using complex numbers.
The puzzle arises from the fact that the physical world
in not a numerical entity and does not contain
numbers. So one must reflect on exactly which
structures of numerical constructions could be
isomorphic to, and hence faithfully represent, physical
structures.

4. The Adage
If we think of certain mathematical constructions
as abstract tools by means of which we represent
the physical structure of the world, then we must
bear in mind a profound saying…
If the only tool you have is a hammer, everything
looks like a nail.
That is, we should strive to construct and evaluate
many mathematical tools, and be sensitive to
when one seems to work better than another.
Perhaps that is because it is better suited to the
problem at hand.

5. Geometry
The Greeks divided mathematics into geometry (the
theory of magnitudes) and arithmetic (the theory of
numbers).
Much of the power of modern science derives from
the use of numerical or algebraic representations of
geometrical structure. This was first made possible via
the introduction of coordinate systems, which
associate points in a geometrical space with sets of
real numbers.
Nonetheless, the most basic geometrical structure is
still described without numbers.

6. Topology
 In order to organize a set of points into a space,
some additional structure must be imposed on them.
 The most fundamental such structure determines
facts about continuity in the space, including the
continuity of functions from one space to another.
This is the topological structure.
 This level of structure is defined without regard to
either distance (metrical structure) or straightness
(affine structure): hence the rubric rubber sheet
geometry.

7. Visually

8. Standard Topology
 The basic notion in the usual formulations
of topology is the open set.
 Every other notion–closed set, connected
space, continuous function, boundary,
compactness, Hausdorff, etc.– is
ultimately defined in terms of the open
set structure.

9. The Architecture of Topology
Open set ⇔ Closed set ⇔ Neighborhoods
Connected
Space
Continuous
Function
Boundary of
a Set
Curve (continuous function
from real line into space)
Path
(image of a curve)

10. Informal Explication
“an open set is one in which every point has
some breathing space” M. Crossley, Essential
Topology
“In topology and related fields of mathematics, a
set U is called open if, intuitively speaking, you
can ‘wiggle’ or ‘change’ any point x in U by a
small amount in any direction and still be inside U.
In other words, x is surrounded only by elements
of U; it can’t be on the edge of U.”–Wikipedia

11. Visually

12. The Axioms
Definition: A topological space is a set, X,
together with a collection of subsets of X,
called “open” sets, which satisfy the following
rules:
T1. The set X itself is “open”.
T2. The empty set is “open”.
T3. Arbitrary unions of “open” sets are “open”.
T4. Finite intersections of “open” sets are
“open”.

13. For Example
Consider a space with only 2 points, p
and q. There are four standard
topologies:
The discrete topology: {p, q}, {p}, {q}, ∅.
The indiscrete topology: {p, q}, ∅.
Two Sierpinski spaces: {p, q} , {p}, ∅ and
{p, q}, {q}, ∅.

14. Why Should This Work?
If this particular mathematical tool—the analysis of the
continuity properties of a space in terms of its open set
structure—is a direct way to describe physical space or
space-time, then there should be some physical feature of
the world that determines which sets of events are open
sets.
It is not obvious what such a physical feature would be.
We could, of course, postulate it as a primitive fact about
sets of events—that some, but not others constitute open
sets—but that should be a last resort.

15. Alternative Geometrical
Primitive: the Line
Rather than the open set, there is a better
fundamental notion upon which a theory of sub-
metrical geometry can be built: the line.
More exactly, the “open” line, in the sense that
both open and closed line segments are “open”
and a circle is “closed”: from any point on the
line one can move continuously to any other
given point, but only by moving in one direction.
An open line in this sense has a structure
represented by a linear order among the points.

16. Visually

17. Theory of Linear Structures
lines
neighborhoods
≠ neighborhoods
open sets
initial-part
open sets
initial-part
closed sets
(=)continuous
functions
≠ continuous
functions
connected space
≠ connected space

18. Linear Orders
A linear order on a set S is a relation, which we
will symbolize by “≥”, that satisfies three
conditions:
 For all p, q, r ∈ S
1) If p ≥ q and q ≥ p, then p = q (Antisymmetry)
2) If p ≥ q and q ≥ r, then p ≥ r (Transitivity)
3) p ≥ q or q ≥ p (Totality)

19. Intervals
An interval in set with a linear order is a
subset of at least two points such that
for any p, q in the subset, all points
between p and q in the order are in the
subset. (Dedekind)

20. Linear Structures (1st
type)
A Linear Structure is a set S together with Λ a
set of subsets of S called the lines of S that
satisfy:
LS1 (Minimality Axiom): Each line contains
at least two points.
LS2 (Segment Axiom): Every line λ admits of
a linear order among its points such that a
subset of λ is itself a line if and only if it is an
interval of that linear order.

21. Linear Structures con’t
LS3 (Point-Splicing Axiom): If λ and µ are lines
that have in common only a single point p that is
an endpoint of both, then λ ∪ µ is a line provided
that no lines in the set (λ ∪ µ) – p have a point in
λ and a point in µ.
LS4 (Completion Axiom): Any linearly ordered
set σ such that all and only the closed intervals in
the order are closed lines is a line.

22. Non-Uniqueness of Order
According to this first set of axioms, every
line can be represented by a linear order
among its points. But evidently there are
two such linear orders that will do the job,
one the inverse of the other. Each will
imply the same intervals, and so the same
structure of segments. (A segment of a line λ is
a subset of λ that is a line.)

23. Lines on a Square Lattice

24. Neighborhoods
A set Σ is a neighborhood of a point
p iff every line with p as an endpoint
has a segment with p as an endpoint
in Σ.

25. Neighborhoods on a Square Lattice

26. Open Sets
A set Σ in a Linear Structure is an open set iff it
is a neighborhood of all of its members.
(NB: this definition looks identical to a
definition that appears in standard topology, with
neighborhood replaced by neighborhood. But in
standard topology, a neighborhood of a point is a
set containing an open set containing the point.)

27. Theorem
The collection of open sets in a
Linear Structure satisfy the axioms
of standard topology, i.e., the open
sets are open sets.

28. Some numbers

29. This suggests….
Evidently many (in an obvious
sense most) standard topologies
on a finite point set cannot be
generated from a Linear Structure
on that set.
I call such topologies
geometrically uninterpretable.

30. 2-point Linear Structures
Discrete topology: open sets
{p,q}, {p}, {q}, Ø
p q
p q
Indiscrete topology: open sets
{p,q}, Ø

31. But…
A little further thought shows this to be incorrect! We
can understand all finite point topologies in terms of
“wiggles”.
In the sort of Linear Structure we have constructed
so far, we have treated the lines as two-way streets:
if a small “wiggle” along a line can take you from p
to q, then a small wiggle along the same line can
take you from q to p.

32. However…
Suppose we treat the lines as one-way streets: to
specify a line one has to specify both a set of
points that constitute it and a direction, i.e., only
one linear order represents a line, not two.
The intuitive notion of a “small wiggle” is a
continuous motion long a line in the direction of
the line.

33. Directed Linear Structures
This gives rise to the notion of a Directed Linear
Structure. The axioms are modified in the obvious
way: all and only the directed intervals in a linear
order are segments of a line, etc.
The Splicing Axiom now requires that to splice two
lines, the point p must be the final point of one
and the initial point of the other.

34. Outward Neighborhoods,
Outward Open Sets
A set Σ is an outward neighborhood of a
point p iff every line with p as an initial
endpoint has a segment with p as an initial
endpoint in Σ.
 A set Σ in a Linear Structure is an outward
open set iff it is an outward neighborhood of
all of its members.

35. Directed LS for Two-Point Space
p q
p q
p q
p q
Outward open sets
{p,q}, {p}, {q}, Ø
{p,q}, {q}, Ø
{p,q}, {p}, Ø
{p,q}, Ø

36. More Numbers
# of points topologies Directed LS Topologies
from DLS
1 1 1 1
2 4 4 4
3 29 64 29
4 355 4,096 355
5 6,942 1,048,576 6,942

37. Theorem
Every finite-point topology is generated
by some finite-point Directed Linear
Structure. Typically, many distinct Linear
Structures give rise to the same
topology, so one loses geometrical
information if one only knows the
topology.

38. Example
These DLSs generate the same topology
(viz. the indiscrete topology).
p q
r
p q
r

39. Geometrically
Uninterpretable Topologies
There are, however, still geometrically
uninterpretable topologies, topologies
generated by no Directed Linear
Structure. They all contain infinitely many
points.

40. The Geometry of a Part of a
Space
In the Theory of Linear Structures, unlike
standard topology, the geometry of a
part of a space is defined in the natural
way: by simple restriction. That is, the
Linear Structure of a part of a space is
given by the lines that are contained in
that part.

41. Space-Time: Why a 4-d
Manifold is Unmotivated
A topological 4-dimensional manifold has an
open set structure that everywhere is locally
isomorphic to a 4-d Euclidean space. From our
point of view, the obvious reason to expect this
would be because the Linear Structure of space-
time is locally isomorphic to the Linear Structure
of 4-d Euclidean space.

42. Lines in Euclidean Space

43. Lines in Newtonian Space-
Time?
time

44. Lines in Relativistic Space-
Time?

45. Wald on “Mixed” Lines
 ”The length of curves which change from timelike to
spacelike is not defined” (General Relativity, p. 44).
So let’s eliminate those curves: the Linear Structure
of a Relativistic space-time is not the same as that of
any Euclidean space.

46. Physics
If the fundamental sub-metrical geometrical
structure is the line, then when we turn to
physics, we should ask: what physical feature
of the universe could generate physical lines?
More generally, what physical feature of the
universe naturally generates a linear order
among the points of space-time?

47. Time
Intuitively, time provides a directed linear
ordering of events. It is the natural place to
look for a source of physical lines.

48. Newtonian and Neo-
Newtonian Space-Time
In Newtonian or Neo-Newtonian space-time, if
one asks for a maximal set of events which is
linearly ordered in time, one gets a set of
points, one at each instant of time. This set of
points will not typically look like any sort of line:

49. Newtonian Space-Time
time

50. Relativistic Space-Time
(Globally Hyperbolic)
In a Relativistic space-time (Lorentzian
pseudo-metric) with no closed time-like
curves (no “time travel”), if one asks for a
maximal set of events which is linearly
ordered in time, what one gets is a
continuous time-like or null curve.
The light-cone structure forces such a set
to intuitively form a line.

51. Relativistic Space-Time

52. From Sean Carroll’s Book

53. Only Trivial Geometry on a
Spacelike Hypersurface
If we only admit timelike-or-null lines, then
when we restrict the geometry to a space-like
hypersurface we get no lines at all: there is no
intrinsic spatial geometry.
Other slices, though, have the expected
Relativistic structure.

54. Linear Structure of Hyperplanes

55. Relativistic Structure is Built
Into the Linear Structure
If we follow this recipe, the light-cone
structure of a space-time is already definable
from its Linear Structure, without use of any
metrical notions.
In particular, a closed line with endpoints p and
q is a straight lightlike line just in case it is the
only closed line with these endpoints. So one
can recover the light-cone structure directly
from the Directed Linear Structure.

56. Recovering the Whole
Relativistic Metric
To get the full Relativistic
(pseudo-)metric, one needs to attribute
a “length” to these lines, i.e. the proper
time along them. This is enough to
determine all the spatio-temporal
structure postulated by Relativity.

57. How the Mathematics
Describes Physics
If we use the Theory of Linear Structures to characterize
the geometry of a space, then the topology is
determined by the directed lines—linearly ordered sets of
points—in the space.
If we accept that time linearly orders events, then the
maximal sets of temporally ordered events form a
natural physical directed linear structure in space-time.
In Relativity—but not classical physics—this turns out to
be just the geometrical structure the physics need. Time
invests space-time with this geometrical structure.